我有按'ID'分组的数据。每个"身份证"在不同的日期有不同的药物。在每次连续运行'drug'时,我希望只保留第一行。这应该按组完成,即在每个"ID"内。数据中显示了两个示例:
ID date drug
1 01/01/2020 A # first row in run 1 of 'A' for ID 1: keep
1 07/01/2020 A # 2nd row in run 1 of 'A' for ID 1: drop
1 09/01/2020 B
1 15/01/2020 A
2 01/02/2020 C
2 13/02/2020 D
2 17/02/2020 C # first row in run 2 of 'C' of ID 2: keep
2 18/03/2020 C # 2nd row in run 2 of 'C' of ID 2: drop
2 19/03/2020 E
所需输出:
ID date drug
1 01/01/2020 A
1 09/01/2020 B
1 15/01/2020 A
2 01/02/2020 C
2 13/02/2020 D
2 17/02/2020 C
2 19/03/2020 E
我已经尝试了以下方法,但我不能使它起作用,因为它会删除那些来自同一组但后来出现的药物,例如它会下降15/01/2020,17/02/2020和18/03/2020,因为它只需要按组进行第一次观察。
df_selection <- df %>%
group_by(ID) %>%
arrange(ID,date) %>%
group_by(ID, drug) %>%
slice(1L) %>%
arrange(ID,date)
我已经尝试了很多组合,但我不能使它工作。我真的很感激你的帮助!
另一个例子来演示一个'ID'中的最后一个'drug'与下一个'ID'中的第一个'drug'相同,这里是drug' B':
ID date drug
1 01/01/2020 A
1 07/01/2020 A
1 09/01/2020 B # first row in a run of 'B' for ID 1: keep
1 15/01/2020 B # 2nd row in a run of 'B' for ID 1: drop
2 01/02/2020 B # first row in a run of 'B' for ID 2: keep
2 13/02/2020 B # 2nd: drop
2 17/02/2020 B # 3rd: drop
2 18/03/2020 E
2 19/03/2020 E
使用data.table
:
setDT(df)[rowid(rleid(drug)) == 1]
# ID date drug
# 1: 1 01/01/2020 A
# 2: 1 09/01/2020 B
# 3: 1 15/01/2020 A
# 4: 2 01/02/2020 C
# 5: 2 13/02/2020 D
# 6: 2 17/02/2020 C
# 7: 2 19/03/2020 E
如果在每个'ID'中考虑'drug'的运行,我们需要…
df[rowid(rleid(ID, drug)) == 1]
…处理以下情况:
ID date drug
1: 1 01/01/2020 A
2: 1 07/01/2020 A
3: 1 09/01/2020 B
4: 1 15/01/2020 B # This 'B' belongs to 2nd run in ID 1
5: 2 01/02/2020 B # This 'B' belongs to 1st run in ID 2
6: 2 13/02/2020 B
7: 2 17/02/2020 B
8: 2 18/03/2020 E
9: 2 19/03/2020 E
df %>% filter(drug != lag(drug, default = ""))
或者,如果您想保留一个ID的药物首次出现,即使它与先前ID的最后一种药物相匹配(例如,假设ID2的第一种药物是a,因此我们想保留它):
df %>%
filter(drug != lag(drug, default = "") |
ID != lag(ID, default = 0))
使用base R
和rle
subset(df, with(rle(drug), !duplicated(rep(seq_along(values), lengths))))
希望此代码适用于您的一般情况
> subset(df, sequence(rle(drug)$lengths) == 1)
ID date drug
1 1 01/01/2020 A
3 1 09/01/2020 B
4 1 15/01/2020 A
5 2 01/02/2020 C
6 2 13/02/2020 D
7 2 17/02/2020 C
9 2 19/03/2020 E